point cloud library - PCL CUDA: How to use pcl::cuda::RandomSampleConsensus

Question

I am trying to use the CUDA RANSAC implementation from PCL version 1.11 from here. I have a simple example that runs without errors when I use storage type of pcl::cuda::Host, but when I try to use the GPU side (pcl::cuda::Device), it fails with error
terminate called after throwing an instance of 'thrust::system::system_error' what(): trivial_device_copy D->H failed: cudaErrorInvalidValue: invalid argument Aborted (core dumped)

How do I actually instantiate and run the CUDA GPU version of pcl::cuda::RandomSampleConsensus?

When I ran the example through the debugger, the debugger crashed but said that it got up to SampleConsensusModel1PointPlane::selectWithinDistance().

running cuda-memcheck gives:

========= Program hit cudaErrorInvalidValue (error 1) due to "invalid argument" on CUDA API call to cudaMemcpyAsync.
=========     Saved host backtrace up to driver entry point at error
=========     Host Frame:/lib/x86_64-linux-gnu/libcuda.so.1 [0x3343d3]
=========     Host Frame:/usr/local/lib/libpcl_cuda_sample_consensus.so.1 [0xa274d]
=========     Host Frame:/usr/local/lib/libpcl_cuda_sample_consensus.so.1 (_ZN3pcl4cuda25SampleConsensusModelPlaneINS0_6DeviceEE20selectWithinDistanceERN6thrust13device_vectorI6float4NS4_16device_allocatorIS6_EEEEifRSt10shared_ptrINS5_IiNS7_IiEEEEER6float3 + 0x8f) [0x3b7cf]
=========     Host Frame:/usr/local/lib/libpcl_cuda_sample_consensus.so.1 (_ZN3pcl4cuda21RandomSampleConsensusINS0_6DeviceEE12computeModelEi + 0x143) [0x50723]
=========     Host Frame:./cuda_random_sample_consensus [0xd220]
=========     Host Frame:/lib/x86_64-linux-gnu/libc.so.6 (__libc_start_main + 0xf3) [0x270b3]
=========     Host Frame:./cuda_random_sample_consensus [0xccde]
=========
========= Program hit cudaErrorInvalidValue (error 1) due to "invalid argument" on CUDA API call to cudaGetLastError.
=========     Saved host backtrace up to driver entry point at error
=========     Host Frame:/lib/x86_64-linux-gnu/libcuda.so.1 [0x3343d3]
=========     Host Frame:/usr/local/lib/libpcl_cuda_sample_consensus.so.1 [0x8b566]
=========     Host Frame:/usr/local/lib/libpcl_cuda_sample_consensus.so.1 (_ZN3pcl4cuda25SampleConsensusModelPlaneINS0_6DeviceEE20selectWithinDistanceERN6thrust13device_vectorI6float4NS4_16device_allocatorIS6_EEEEifRSt10shared_ptrINS5_IiNS7_IiEEEEER6float3 + 0xa1) [0x3b7e1]
=========     Host Frame:/usr/local/lib/libpcl_cuda_sample_consensus.so.1 (_ZN3pcl4cuda21RandomSampleConsensusINS0_6DeviceEE12computeModelEi + 0x143) [0x50723]
=========     Host Frame:./cuda_random_sample_consensus [0xd220]
=========     Host Frame:/lib/x86_64-linux-gnu/libc.so.6 (__libc_start_main + 0xf3) [0x270b3]
=========     Host Frame:./cuda_random_sample_consensus [0xccde]
=========
terminate called after throwing an instance of 'thrust::system::system_error'
  what():  trivial_device_copy D->H failed: cudaErrorInvalidValue: invalid argument
========= Error: process didn't terminate successfully
========= No CUDA-MEMCHECK results found

Code that produces error:

src/main.cu

#include <iostream>
#include <pcl/cuda/sample_consensus/ransac.h>
#include <pcl/cuda/sample_consensus/sac_model_plane.h>

int main() {
    pcl::cuda::PointCloudAOS<pcl::cuda::Device> ground_cloud;

    ground_cloud.width = 50;
    ground_cloud.height = 50;
    ground_cloud.is_dense = false;

    pcl::cuda::PointXYZRGB point;

    for (int i = 0; i < ground_cloud.width * ground_cloud.height; i++) {

        point.x = 5.;
        point.y = 5.;
        point.z = 5.;

        point.rgb.r = 0.5;
        point.rgb.g = 0.5;
        point.rgb.b = 0.5;

        ground_cloud.points.push_back(point);
    }

    pcl::cuda::SampleConsensusModelPlane<pcl::cuda::Device>::PointCloudPtr 
        cloudPtr(new pcl::cuda::SampleConsensusModelPlane<pcl::cuda::Device>::PointCloud (ground_cloud));

    pcl::cuda::SampleConsensusModelPlane<pcl::cuda::Device>::Ptr 
        plane(new pcl::cuda::SampleConsensusModelPlane<pcl::cuda::Device>(cloudPtr));

    pcl::cuda::RandomSampleConsensus<pcl::cuda::Device> ransac(plane);

    ransac.setDistanceThreshold(0.07);
    ransac.setMaxIterations(50);

    bool status = ransac.computeModel();
    
    return 0;
}

CMakeLists.txt:

cmake_minimum_required(VERSION 2.8 FATAL_ERROR)

project(random_sample_consensus)

find_package(PCL 1.2 REQUIRED)

enable_language(CUDA)
set(CMAKE_CUDA_COMPILER nvcc)

include_directories(${PCL_INCLUDE_DIRS} /usr/local/cuda/include)
link_directories(${PCL_LIBRARY_DIRS})
add_definitions(${PCL_DEFINITIONS})


add_executable (ransac src/main.cu)
target_link_libraries (ransac ${PCL_LIBRARIES})

add_compile_options(-w)

question from:https://stackoverflow.com/questions/65912893/pcl-cuda-how-to-use-pclcudarandomsampleconsensus

Answer 1

Waitting for answers